Topic: Hacking the S7 - improving efficiency through minor hardware manipulation - page 7. (Read 26495 times)

I have over the Years bought several PicKit 2 & 3, all clones from China and not had a single problem. If you can put up with the delivery time then you can pick up a Pickit3 for $12 including shipping.


Rich

Wow, I wish I had known that when I ordered. I guess I should not complain about a 5% difference.

If you happen to have an edu email address you can also get a 25% student / academic discount from microchipDIRECT. My order was just under $48 shipped for the PICkit 3 + the Programming Cable Kit.

You can find it cheaper but shipping was fast and that price works for me.

I posted my recommendation for getting the PICkit3 direct from the manufacturer Microchip Technology Inc. here: https://bitcointalksearch.org/topic/m.15332015

The pricing is not bad considering you know it is not a gray market item. The company only offers overnight shipping to the United States and it costs about $7. I think my total cost for the PICkit3, extension cable, and overnight shipping was ~$50.

it could be this..

i know some pools, like zpool the temps on the board drop down about 5c.. i think its because the pool switches coins very often.



if they go under around 53c they drop out and cgminer needs to be restarted to turn back on..

i made a sh script to check for that.. sucks you have to run it every time you reboot the miner, but it runs until you shut it down.

Share diff only affects what shares are sent to the pool and has no effect on the work being done. The chips idle when work is not sent from the server and cgminer has nothing to feed them with. Bitmain chips (at least up through BM1384) return all shares diff 1 and higher and cgminer filters out the results based on what the pool is asking for.

Along these lines, just how much does ShareDiff affect actual 'work' (power used) load? If the dif is not too hard and share rate is too low do the ASIC's spend time idling and just twiddling their thumbs so to speak?

I know that Antpool samples throughput every 1/2hr and as a result ShareDiff bounces up and down pretty wildly whereas on Kano it changes every minute or so. On Antpool the diff would often go up down 25% or more and stay there for the 1/2hr, Kano, just a few % each change. Naturally, more samples in a shorter time translates to better and more accurate averaging response time...

Awesome thread keep posted thread loke this

I could see it being an issue from mining starting and stopping instead of being fed continuous work, if the boards cool down and drop out before mining kicks back up. That shouldn't be affected by share diff but by availability of work.

Possibly related to most pools limiting the share rate each miner gets? Done to smooth out the pool servers data loading and keep it reasonably fair for the wide range of miner speeds folks are using. I *think* most try and keep it around 8 shares/sec by adjusting the ShareDiff higher/lower as needed to average that share rate, no? That in turn limits how much 'work' the miner is doing at any given time.

When Solo mining I'd hope the only throttle is the miner vs pool access fairness...

i got this one
http://www.aliexpress.com/item/PICKIT3-Programmer-PIC-ICD2-PICKit-2-PICKIT-3-Programming-Adapter-Universal-Programmer-Seat-FZ0508-Free-Shipping/1967812035.html?spm=2114.13010608.0.54.j30t2e

it took a few days to get here to usa.. about 10.. but it worked..

keep in mind the pool makes a huge difference on the power on the s7..
while i can run kanos pool with 640s on all boards.. i cant solo mine or mine peercoin without bumping mine up to 680


i dunno why..

For reasons which have probably been explained extensively, it only works with 135-chip S7 and not 162-chip S7 because you're adjusting the output voltage of the main regulator on the board and with 162-chip miners there is no main regulator to adjust. The same applies to the S5.

Really appreciate this thread thanks! There seem to be different versions sellers of the pickit3. As a noob is there a standard I should order? expect to pay about $65? Also this thread suggest this mod will only work with the 135 chip BM miner. I will try the s7, but if I'm going after an S5 mod I should look elsewhere?

i got my pickit3 today and i modded my other s7..
so far i have it running at 550 with 660 on all 3 boards. im at 956w @ 3700 ghs..
im using the s5 fan as intake and a afb1212she has exhaust.


looking at the back does anyone know how the boards are numbered??

1-2-3 or 3-2-1?

i can run them at 650 but board number 1 drops out after awhile and goes to 30c. it did turn back on after about 5 minutes..

at 640 i start to get x's all over.

edit..

these things love the heat.. i lowered the fans down to 60% and the temps are now upper 50s to mid 60s.. sucks on automatic they are pretty much full speed..
my watt usage went down to 848-850, my speed is right at 3700.. errors are 0003-0005%

that puts me at .23j/ghs

i still think i can drop a couple boards down to 650.. maybe ill try that tomorrow.. not sure how much more watt savings tho 10mv would be..

Maybe half the S7LN and pretty much none of the S7 that I've worked up have had the same minimum stable voltage for any given frequency. Since each board has so many chips on it, in series, probably randomly scattered with regard to per-chip performance, there are too many variables to expect uniform performance at the edge cases.

Good results though. 900W for 3.9TH is pretty solid.

I wanted to update my results on my modded S7.  So a little disappointed that my 3 boards each needed a different mV setting to run stable.  One at 630, one at 640 and one at 650 mV.  But they seem to work out now OK.  I was able to program the boards even though they had soldered holes by firmly pushing the pins to the soldered holes and after a while was able to get them to program (3.55 V worked for me).  Here's the results.  With 5 additional case fans and a fan controller and a ASUS wireless AP the S7 running at 562.5M was pulling about 919 watts on an EVGA 1600w T2 Titanium (so that would be closer to 900w without the additional equipment).  Stock runs about 1,175w at 562.5M, so that's about 270w less (-23%).  At 562.5M it pulls about 3.9 TH/s with 0.003-0.006% HW errors.  The fans running at 33% manual = 48 on two outer boards and 56 on the middle board temps. (C).  So it would have been nice if all the boards would have been stable at the same mV setting, it still came out pretty well.  Well worth the price of the PickIt 3 programmer and the effort put into it.  Thanks Sidehack.
 

The hacked S7-LN from the group buy is working nicely, thanks again Sidehack and Phil!  

I have now some measurements for you Phil.
The PSU I'm using is Super Flower Golden Green SF-500P14XE 500W.
The meter used for both voltage and current, with my test setup is HT Instruments HT7011 clamp meter.
Voltage is approx. 222 VAC
Current is approx. 2.0 A
So the power at the wall is approx. 444 W

Not bad.  

It is almost the same as you measured Phil (442W).

I have tried the 1 Fan Rom and it runs fine with a single fan.


Rich

It appears that I was able to program the 1 board without the soldered holes, but only after I increased the voltage from the programmer from the VDD: 3.3 v to 3.55 v.  Now to test it.

i picked up one of these modded s7s from sidehack..
with the stock fans i was having trouble keeping it running at night.. even lowering the fans to 20% was not enough to keep the boards warm enough.. once they got to 40c or so they will shut off.

we are having pretty good weather here, but it still cools off at night.. upper 50s at night and 80s during the day.


i put on a pair of AFB1212-SHE fans i got for my s5s.
https://www.amazon.com/Delta-Electronics-AFB1212SHE-120x120x38mm-connector/dp/B004Y1HLA8

these fans do have a bit more of a whine to them than the stock fans.. but the air flow is slower and the s7 doesnt whistle, over all i think they are more quiet. like a s5.

running them at AUTO they seem to be running pretty good at 600 speed. right at 4000 ghs and using 1023 at the wall. my unmodded s7 at 600 speed uses 1280 watts.
they speed up to about 3400 rpm during the day and slow down to around 3100 at night. they max out at 3600 so there is still room for when it gets hotter.
the temps are holding right at 55-64c.



i think you can flash the 1-fan rom on it and just use one stock fan.. not sure tho.. i didnt try. i might try it on my other s7 when i get the pickit3 i ordered from china.


overall the mod is well worth it.. you dont lose much speed (actually i was running my s7 at 600 already) and you lose almost 300 watts. you do have to do something with the fans tho, the boards need to be above 50c to stay stable.